CD19 Alterations Emerging after CD19-Directed Immunotherapy Cause Retention of the Misfolded Protein in the Endoplasmic Reticulum

Abstract

We previously described a mechanism of acquired resistance of B-cell acute lymphoblastic leukemia to CD19-directed chimeric antigen receptor T-cell (CART) immunotherapy. It was based on in-frame insertions in or skipping of CD19 exon 2. To distinguish between epitope loss and defects in surface localization, we used retroviral transduction and genome editing to generate cell lines expressing CD19 exon 2 variants (CD19ex2vs) bearing vesicular stomatitis virus G protein (VSVg) tags. These lines were negative by live-cell flow cytometry with an anti-VSVg antibody and resistant to killing by VSVg-directed antibody-drug conjugates (ADCs), suggestive of a defect in surface localization. Indeed, pulse-chase and α-mannosidase inhibitor assays showed that all CD19ex2vs acquired endoplasmic reticulum (ER)-specific high-mannose-type sugars but not complex-type glycans synthesized in the Golgi apparatus. When fused with green fluorescent protein (GFP), CD19ex2vs (including a mutant lacking the relevant disulfide bond) showed colocalization with ER markers, implying protein misfolding. Mass spectrometric profiling of CD19-interacting proteins demonstrated that CD19ex2vs fail to bind to the key tetraspanin CD81 and instead interact with ER-resident chaperones, such as calnexin, and ER transporters involved in antigen presentation. Thus, even the intact domains of CD19ex2vs cannot be easily targeted with ADCs or current CD19 CARTs but could serve as sources of peptides for major histocompatibility complex (MHC)-restricted presentation and T-cell receptor (TCR)-mediated killing.

Keywords: RNA splicing; immunotherapy; membrane proteins; membrane transport; protein folding.

FIG 1

Mutations in CD19 exon 2 result in cell surface localization defects. (A) Schematic of VSVg-CD19 FL, VSVg-105R, and VSVg-Δex2 retroviral constructs. The VSVg tag was inserted in exon 1 right after the signal peptide sequence to prevent cleavage of the epitope. (B) Western blotting with anti-VSVg antibody in transduced Nalm6 cell lines. (C) Flow cytometry of live or fixed/permeabilized transduced cell lines stained with anti-VSVg–FITC or anti-CD19 (FMC63)–PE antibody. (D) Flow cytometry of live transduced Nalm6 cell lines grown at 37°C or 32°C for 72 h before staining with anti-VSVg–FITC for 1 h. The cells were then washed with PBS as a control or glycine to dissociate any surface-bound antibody. (E and F) Killing assays using anti-VSVg antibody and anti-mouse antibody–MMAF 2-ADC in transduced Nalm6 cell lines grown at 37°C (E) or 32°C (F) for 72 h.

FIG 2

CD19ex2 variants exhibit altered glycosylation patterns. (A) Schematic showing cellular compartments where sequential steps of protein glycosylation and processing occur. (B) Schematic showing the intended targets of cleavage by endo H and endo F glycosidases, as well as the α-mannosidase inhibitor swainsonine. (C) Western blot with anti-CD19 or antiactin antibody of protein extracts from the transduced Nalm6 cell lines after treatment with DMSO or 10 μg/ml swainsonine for 24 h. (D) Western blot with anti-CD19 or antiactin antibody of lysates from the transduced 697 cell lines following treatment with DMSO or 10 μg/ml swainsonine for 24 h. (E) Western blot with anti-CD19 or antiactin antibody of protein lysates from the transduced Nalm6 cell lines following in vitro mock or PNGase F treatment. (F) Western blot with anti-CD19 or antiactin antibody of protein lysates from the transduced 697 cell lines following in vitro mock or PNGase F treatment. (G) Nalm6-ΔCD19 cells or cells transduced with CD19-FL or CD19ex2vs constructs were radiolabeled for 15 min, chased for 1 or 2 h, and immunoprecipitated using a monoclonal antibody against human CD19. Immunoprecipitates were treated with endo H (H) or PNGase F (F) before analysis on an SDS-PAGE gel. CHO, high-mannose-type glycans; CHO*, complex-type glycans; NAG, N-acetylglucosamine.

FIG 3

Inhibition of CD19 glycosylation does not affect its subcellular localization. (A) Immunofluorescence of Nalm6-ΔCD19 cells expressing FL-CD19–GFP (top schematic) after treatment with DMSO or swainsonine for 24 h. The plasma membrane was stained with wheat germ agglutinin-Alexa Fluor 647 (converted to red), and nuclei were stained with DAPI (blue). The histograms show colocalization of CD19-GFP (green) and plasma membrane (red) channels. (B) Pearson's correlation colocalization analyses for FL-CD19–GFP and plasma membrane. Three separate fields containing at least 100 cells were analyzed for each condition. The error bars indicate standard deviations. (C) Immunofluorescence of 697-ΔCD19 cells expressing CD19-FL-GFP after treatment with DMSO or swainsonine for 24 h. The plasma membrane was stained with wheat germ agglutinin-Alexa Fluor 647 (converted to red), and nuclei were stained with DAPI (blue). The histograms show colocalization of CD19-GFP (green) and plasma membrane (red) channels. (D) Pearson's correlation colocalization analyses of FL-CD19–GFP and plasma membrane. Three separate fields containing at least 100 cells were analyzed for each condition. (E) Immunofluorescence of Nalm6-ΔCD19 cells transduced with N86A-CD19–GFP fusion protein (top schematic). The plasma membrane was stained with wheat germ agglutinin-Alexa Fluor 647 (red).

FIG 4

CD19ex2 variants localize in the endoplasmic reticulum. (A) Schematic of the CD19-FL, CD19Δex2, and CD19-105R–GFP fusion protein constructs expressed in Nalm6-ΔCD19 cells. (B) Western blot with anti-CD19 antibody of lysates from the indicated cell lines. (C) Live-cell flow cytometry for empty (black), CD19-FL (blue), CD19-Δex2 (green), and CD19-105R (red) using anti-CD19–PE antibody. (D) Immunofluorescence confocal microscopy of GFP construct-transduced cell lines (green). The plasma membrane was stained with wheat germ agglutinin-Alexa Fluor 647 (converted to red). The endoplasmic reticulum was stained with anticalnexin (Cell Signaling)/anti-rabbit antibody–Alexa Fluor 594 (converted to magenta). (Right) Histogram localization analysis showing overlaps of channels. (E) Pearson's correlation coefficient analyses showing strong colocalization of plasma membrane and CD19-FL and endoplasmic reticulum and CD19ex2vs isoforms. The error bars indicate standard deviations.

FIG 5

Disruption of the CD19 Cys38-Cys97 disulfide bond leads to loss of surface expression. (A) Schematic representation of the CD19 protein indicating the exon 2 residues involved in bisulfite bonds and loop formation and in glycosylation. Depictions of the C97A and N86A/C97A-CD19 protein constructs are also shown. (B) i-Tasser protein structure prediction of CD19-FL (top) and CD19-105R (bottom). For simplicity, only exons 1 and 2 are shown. (C) Western blot with anti-CD19 or antiactin antibody of protein extracts from the transduced Nalm6 cell lines following treatment with DMSO or 10 μg/ml swainsonine for 24 h. Various degrees of CD19 protein glycosylation are shown by gel shift of bands. CHO, high-mannose-type glycans; CHO*, complex-type glycans; NAG, N-acetylglucosamine. (D) Western blot with anti-CD19 or antiactin antibody of protein lysates from the transduced Nalm6 cell lines following in vitro mock or PNGase F treatment. (E) Western blot with anti-CD19 or antiactin antibody of protein lysates from the transduced 697 cell lines following in vitro mock or PNGase F treatment. (F) Live-cell flow cytometry using anti-CD19–PE antibody of transduced Nalm6 and 697 cell lines.

FIG 6

Disruption of the CD19 Cys38-Cys97 disulfide bond leads to endoplasmic reticulum retention. (A) Immunofluorescence confocal microscopy of the indicated CD19-GFP construct (green)-transduced Nalm6 cell lines. The plasma membrane was stained with wheat germ agglutinin-Alexa Fluor 647 (converted to red), the endoplasmic reticulum was stained with anticalnexin (Cell Signaling)/anti-rabbit antibody–Alexa Fluor 594 (converted to magenta), and nuclei were stained with DAPI (blue). (Right) Histogram localization analysis showing overlap of CD19-GFP, ER/calnexin, and plasma membrane channels. (B) Pearson's correlation colocalization analyses of green (CD19) and red (plasma membrane) channels or green (CD19) and ER/calnexin channels for the indicated Nalm6 cell lines. Three separate fields containing at least 100 cells were analyzed for each condition. The error bars indicate standard deviations. (C) Immunofluorescence confocal microscopy of the indicated CD19-GFP construct (green)-transduced 697 cells. The plasma membrane was stained with wheat germ agglutinin-Alexa Fluor 647 (converted to red), and nuclei were stained with DAPI (blue). (Right) Histogram localization analysis showing overlap of CD19-GFP (green) and the plasma membrane (red). (D) Pearson's correlation colocalization analyses of green (CD19) and red (plasma membrane) channels for the indicated 697 cell lines. Three separate fields containing at least 100 cells were analyzed for each condition.

FIG 7

Endogenous CD19ex2 variants generated by genome editing are also retained in the endoplasmic reticulum. (A) Schematic of CRISPR/Cas9 mutant clones to model CD19-Δex2 (clone 11) or CD19-105R (clone 12). Shown is a Western blot with anti-CD19 or anti-GAPDH antibodies to compare protein expression between CRISPR/Cas9 and retroviral overexpression cell line models. (B) Flow cytometry of live or fixed and permeabilized Nalm6 parental and CRISPR/Cas9-modified cells expressing endogenous CD19 wild-type (parental), CD19-Δex2 (clone 11), and CD19-105R-like (clone 12) isoforms. (C) Killing assay with anti-CD19 antibody and anti-mouse DMDM 2-ADC using parental Nalm6 (CD19-WT), ΔCD19 (null), clone 11 (CD19-Δex2), or clone 12 (CD19-105R-like) cells. The error bars indicate standard deviations. (D) Nalm6-ΔCD19, parental, clone 11, and clone 12 cell lines were radiolabeled for 15 min, chased for 1 or 2 h, and immunoprecipitated using a monoclonal antibody against human CD19. Immunoprecipitates were treated with endo H or PNGase F before analysis on an SDS-PAGE gel. CHO, high-mannose-type glycans; CHO*, complex-type glycans; NAG, N-acetylglucosamine. (E) Immunoblotting with anti-CD19 or anticalnexin antibodies following immunoprecipitation with anti-CD19 antibody in the indicated cell lines. (F) Immunofluorescence confocal microscopy of parental Nalm6 (CD19-WT), clone 11 (CD19-Δex2), and clone 12 (CD19-105R-like) cells. The cells were stained with anti-CD19 (3G7; Origen)/anti-mouse antibody–Alexa Fluor 488 (green). The plasma membrane was stained with wheat germ agglutinin-Alexa Fluor 647 (converted to red), the endoplasmic reticulum was stained with anticalnexin (Cell Signaling)/anti-rabbit antibody–Alexa Fluor 594 (converted to cyan), and nuclei were stained with DAPI (blue). (Right) Histogram localization analysis showing overlap of color channels.

FIG 8

CD19ex2 variants preferentially interact with ER/MHC-related proteins. (A) Immunoprecipitates of CD19 from lysates of transduced Nalm6 cell lines were subjected to mass spectrometry. The bars represent normalized total spectra for the identified protein targets. (B) Heat map showing proteins from the Reactome Adaptive Immune System gene set that exhibited differential binding to CD19, CD19-105R, and CD19-Δex2, as determined by mass spectrometry. (C) Double-stained live-cell flow cytometry using anti-CD19–PE and anti-CD81–FITC antibodies on Nalm6 parental and Nalm6-ΔCD81 cell lines. (D) Immunoblotting of CD81, calnexin, calreticulin, UGCGL1, and PDI following immunoprecipitation with anti-CD19 antibody in lysates of the transduced cell lines. The CD19/CD81 double-KO Nalm6 cell line was used as a negative control. (E) Western blot for total levels of CD81, P97, and actin in lysates from the transduced cell lines. The CD19/CD81 double-KO Nalm6 cell line was used as a negative control. (F) Immunoblotting of CD81 following immunoprecipitation with anti-CD19 antibody in lysates from primary patient samples expressing CD19-WT (101 and 105) and CD19-105R immunotherapy-resistant isoforms.